I have recently switched the OS on my laptop from Mandrake 8.0 to Redhat 8.0, and the game I’ve been woking on goes from 17 frames/s to 10 frames/s
Has anyone else seen these problems?
I’m continuing to try and find the source of the problem – I believe that it might be the compiler (Rh->gcc3.2 Vs Md->gcc->2.96)
I’ve also found some strange slowdowns from I/O. I used to have some debugging messages that went allong with each mouse move. On the new RH system these printf’s caused the gl-updates to crawl. Removing the printf’s that went along with the mouse moves brought the performance back to normal (or at least to 10 frames/s).
Any other simmilar experiences with the newest RH?
By the way I’m not trying to start a distro flame war, just trying to work through some problems – all of the distro’s have their own set of problems
Originally posted by amendol: I believe that it might be the compiler (Rh->gcc3.2 Vs Md->gcc->2.96)
And what, exactly, are you basing that belief on? Millions of things changed when you went from Mdk to RH, including video drivers, X version, kernel version, etc. etc., not just your compiler version.
Most likely you have run into the 386 vs 586 problem. All RH packages are compiled without Pentium optimisations, whereas Mdk packages are specifically compiled with such optimisations.
And what, exactly, are you basing that belief on? Millions of things changed when you went from Mdk to RH, including video drivers, X version, kernel version, etc. etc., not just your compiler version.
-Well…
About 6 Months ago I was experimenting with switching to gcc-3.0.4 under Mandrake, and I saw a very simmilar slowdown.
-yes there are many factors, and it could be alot of things including the one that you mentioned. I was asking if anyone had a similar experience and found the problem.
-The NVidia drivers are the same. Still using agp 4x, no fast writes or side band etc.
-X is at a point in the 4x version tree that I expect things to get faster not slower
Most likely you have run into the 386 vs 586 problem. All RH packages are compiled without Pentium optimisations, whereas Mdk packages are specifically compiled with such optimisations.
-I seem to have gotten things to behave a little more normal (about 15.5 fps without compiler optimization flags set). Unfortunately I’m not sure what has brought the speed up. These are the things that I did:
-gcc3.2 gives lots of warnings about the STL and templates in general that gcc2.96 did not. – I managed to fix all the warnings.
-RH 8, has a utility to tweek the kernel’s vm so malloc does not check if the system has enough physical memory. I set it to “Overcommit Memory”
-I don’t believe that I did anything else. I’ll have to try and retrace my steps to see if I can find where the slowdown happened. CVS is your friend.
I just installed RedHat 8.0 and Glut 3.7 on my box. I have run it with both a GeForce 2 MX 400 PCI card and a GeForce 3 Ti500 AGP card and all I can say is “Holy Chiutt!!”. Man this thing flies relative to WinXP (same box).
I believe that your problem is a simple one … the OpenGL drivers installed with RH8.0 DO NOT SUPPORT HARDWARE ACCELERATION. You have to get HDW accelerated drivers from your graphics card vendor.
If you are using an nVidia board then go to the following link and follow the instructions verbatim. It is easy and works like a champ :
Like I said in my previous post, the slowdown has gone away. I’ve managed to get the frame rate up to the same level that it was under Mandrake with the same drivers. I’ve had lots of other things to work on with the engine (getting height map collision detection to work…), and that’s what I’m really interested in, so that’s what I spend my (limited) time on. I am a huge Linux fan and I do all of my work under Linux, but over the years, I’ve gotten a little tired of configuring new distributons, so I tend to leave that to the distributions.I’ll just do the minimum that it takes to get the setup where I want it. I spent some time configuring the old Mandrake system, and I verified that the speed of the OpenGL drivers was at least as good as the Windows drivers. That was my target. Now the Redhat distribution is performing at the same rate, so I’m happy
Currently I suspect that the slowdown was related to variations in the STL and Template implementations between gcc2.x on mandrake and gcc3.2 on redhat. I haven’t confirmed this yet, but the speedup seemed to occur when I fixed some template and STL warnings that I got when compiling under Redhat.
When I get tired of implementing height map collisions in my engine, I’ll re-introduce the Template/STL code that triggered the warnings under redhat, and see if this slows the rendering down. --Then I’ll know if that is what really caused the problem.
STL does not sound like the most obvious
reason for a slowdown, but if you can
somehow prove it with a simple code
snippet, it would be of great interest
indeed.
Originally posted by nigels:
[b]STL does not sound like the most obvious
reason for a slowdown, but if you can
somehow prove it with a simple code
snippet, it would be of great interest
indeed.
Cheers,
Nigel Stewart[/b]
Finished with my height map collision detection
I put some of the code that was causing stl/template warnings back in, and found no change in frame rate. This gives me nothing conclusive.
I’ll have to see if I can retreve the version from cvs that was slower on redhat, and see what change it takes to speed it back up…
